WeRateDog Data Analysis and Visualization Report.¶

By: Benjamin Adekunle Ojo.¶

Introduction:¶

>

The aim of this report is to communicate findings gotten from analysing the weratedog dataset, and interpreting visualization used in arriving at our conclusions.

In [7]:
# Import pakages:

# Data Manipulation Pakages.
import pandas as pd 
import numpy as np

# Data Visualization Pakages. 
import matplotlib.pyplot as plt
import seaborn as sns 

%matplotlib inline 

# Other pakages. 
from IPython.display import Image
In [8]:
Image('weratedog_tweeter_acc.png')
Out[8]:
In [2]:
# Loading Weratedog cleaned dataset

weratedog_data = pd.read_csv('weratedog_tweets_master.csv')
In [3]:
weratedog_data.head()
Out[3]:
tweet_id timestamp source text expanded_urls rating_numerator rating_denominator name dog_stage favorite_count ... confid_probability_1 image_dog_1 Prediction_2 confid_probability_2 image_dog_2 Prediction_3 confid_probability_3 image_dog_3 confidence_max breed_type
0 892177421306343426 2017-08-01 00:17:27 twitter.com/download/iphone This is Tilly. She's just checking pup on you.... twitter.com/dog_rates/status/89217742130634342... 13 10 Tilly NaN 29321 ... 0.323581 True Pekinese 0.090647 True papillon 0.068957 True 0.323581 Chihuahua
1 891815181378084864 2017-07-31 00:18:03 twitter.com/download/iphone This is Archie. He is a rare Norwegian Pouncin... twitter.com/dog_rates/status/89181518137808486... 12 10 Archie NaN 22042 ... 0.716012 True malamute 0.078253 True kelpie 0.031379 True 0.716012 Chihuahua
2 891327558926688256 2017-07-29 16:00:24 twitter.com/download/iphone This is Franklin. He would like you to stop ca... twitter.com/dog_rates/status/89132755892668825... 12 10 Franklin NaN 35308 ... 0.555712 True English_springer 0.225770 True German_short-haired_pointer 0.175219 True 0.555712 basset
3 891087950875897856 2017-07-29 00:08:17 twitter.com/download/iphone Here we have a majestic great white breaching ... twitter.com/dog_rates/status/89108795087589785... 13 10 None NaN 17803 ... 0.425595 True Irish_terrier 0.116317 True Indian_elephant 0.076902 False 0.425595 Chesapeake_Bay_retriever
4 890971913173991426 2017-07-28 16:27:12 twitter.com/download/iphone Meet Jax. He enjoys ice cream so much he gets ... gofundme.com/ydvmve-surgery-for-jax,https: 13 10 Jax NaN 10361 ... 0.341703 True Border_collie 0.199287 True ice_lolly 0.193548 False 0.341703 Appenzeller

5 rows × 24 columns

Analysis and Visalization¶

Upon completing cleaning the dataset and exporting the cleaned data, I moved to analysing my data for possible insight >research questions, of which we analysis using data tables and visualizations, and the following are questions we analysis;

  1. What is the relationship between favourite_count and retweet_count?
  2. Does there exist a ralationship between rating and favourite_count ?
  3. What breed of dogs have the highest rating, retweet, and favourite_count ?

1. What is the Relationship Between favourite_count and retweet_count¶

For this analysis we will be looking at the relationship that exist betweet favourite count of a tweet and it retweet rate. Getting an insight on how much a tweet is retweeted base on the rate at which it is like by views. for this analysis we will be using the following data:

  • favourite_count
  • retweet_count
In [12]:
Image('weratedog_pic_2.png')
Out[12]:

Visualization:¶

In [10]:
# Calculating the correlation between favorite_count and retweet. 
corr_df = weratedog_data[['favorite_count', 'retweet_count']]

corr = corr_df.corr()

# Viewing the correlation in tabluar formate
corr.style.background_gradient()
Out[10]:
  favorite_count retweet_count
favorite_count 1.000000 0.927102
retweet_count 0.927102 1.000000
In [9]:
# Plotting Scatter plot for a better visualization

# Graph Size 
plt.figure(figsize = (20, 15))

# Plotting the scatter plot 
weratedog_data.plot(kind= 'scatter', x= 'retweet_count', y = 'favorite_count', alpha=0.5, c = 'red')

# setting annotation 
plt.ylim((0, 100000))
plt.xlim((0, 40000))
plt.ylabel('Favorite Count')
plt.xlabel('Retweet Count')
plt.title('Relationship between Favorite Count & Retweet Count ')
plt.show()

<Figure size 1440x1080 with 0 Axes>

Findings:¶

From our analsys and visalization above we can observe that there seem to be a positive relationship between the rate at which a tweet is liked by followers and the rate at which it is retweeted, although this confirms our initial premise that tweets with high favourite count are also more likely to have a high retweet count. The correleation rate computed shows that there is a stronge correlation of 86% percent.

To highlight the point we got from computing the correlation, we plotted a scatter plot showing the relationship betwrite ceen favorite count and retweet count. As expected it shows a positive relationship between the rate favourite count and retweet, but we also discovered that not all retweet with a high count also have high favorite count and this make sense in that not all tweet that have hight retweet and can a high favo

2. Does there exist a ralationship between dog_rating and favourite_count?¶

From our previous insight we discovered there is a postitive relationship between the rate at which people like tweets and the how much they share it to other by retweeting the inital tweet, but this was to just establish that there is a relationship between tweets like and retweet rate.

In this new insighful analyssi we will be looking at the relationship between the dog rating and the favourite count of each dog. Understanding this relationship will help us know how fair the rating by people is to what people feel about the dog in the tweets. For this analysis we will be making use of the following data:

  • favorite_count
  • dog_rating
In [20]:
Image('weratedog_pic_3.png')
Out[20]:

Visualization:¶

In [21]:
# Weratedog Data subset. 
dog_fvra = weratedog_data[['tweet_id', 'favorite_count', 'rating_numerator', 'retweet_count']]
In [22]:
# Removing dog rating that are greater than 15.00
# As this data set are outlier, and removing them woundn't significantly affect
# the size of the 

dog_fvra = dog_fvra[dog_fvra['rating_numerator'] <= 15.99]
In [23]:
# Identifing the correlation between dog rating and favorite count
fvra_corr_df = dog_fvra[['rating_numerator', 'favorite_count']]

fvra_corr = fvra_corr_df.corr()

=fvra_corr.style.background_gradient()
Out[23]:
  rating_numerator favorite_count
rating_numerator 1.000000 0.414709
favorite_count 0.414709 1.000000
In [26]:
# Plotting Scatter Plot

# Plot Size
plt.figure(figsize=(10,10), dpi=250)

# Scatter Plot 
dog_fvra.plot(kind = 'scatter', x = 'rating_numerator', y = 'favorite_count', alpha = 0.5, color = 'b')

# Plot Annotation
plt.ylim((0, dog_fvra.favorite_count.max()+ 3))
plt.xlim((0, dog_fvra.rating_numerator.max()+3))
plt.ylabel('Favorite Count')
plt.xlabel('Dog Rating')
plt.title('Dog Rating VS Favorite Count')
plt.show()

<Figure size 2500x2500 with 0 Axes>

Findings:¶

>

Our analysis turns out to be very interesting from correlation and plot, we can obserse that there seems to be little correlation between this data point meaning that neither the rating nor the favorite count is dependent on the other but there still a positive relationship between the two data because the correlation value of 0.41 or 41% is low compare to our expection. This can also mean there is a bias with the rating system, although this is an assumption it gives room for further analysis.

The scatter plot we generated show us that there is a positive relationship between the ratings and favorite count, this goes accouding to our expection and the correlation values.

3. What breed of dogs have the highest rating, retweet, and favourite_count ?¶

>

This analysis aim at understanding what breed type people are most likly to get more retweet, rating, like count. Understating this can help us in predicting which breed of dogs are most likly to get a certain amount of likes , retweets, and rating. To conduct this anlysis we will be using Bar chars and the group by function in pandas. The following dataset would be used in this analysis and they are:

  • Favorite Count
  • retweet count
  • dog_breed
  • dog_rating
In [35]:
Image('weratedog_pic_1.png')
Out[35]:
In [27]:
# Weratedog dataset subset. 
dog_breed = weratedog_data[['tweet_id', 'favorite_count', 'retweet_count', 'rating_numerator', 'breed_type']]
In [28]:
# Removing dog rating that are greater than 15.00
# As this data set are outlier, and removing them woundn't significantly affect
# the size of the 

dog_breed = dog_breed[dog_breed['rating_numerator'] <= 15.99]
In [29]:
# What breed of dogs are the most common 

most_common = dog_breed.groupby(['breed_type'])['tweet_id'].count().sort_values(ascending = False).head(20)

most_common
Out[29]:
breed_type
golden_retriever            129
Labrador_retriever           88
Pembroke                     85
Chihuahua                    76
pug                          53
chow                         40
Samoyed                      37
Pomeranian                   37
toy_poodle                   36
malamute                     29
cocker_spaniel               27
French_bulldog               25
Chesapeake_Bay_retriever     23
miniature_pinscher           21
German_shepherd              20
Siberian_husky               19
Maltese_dog                  18
Eskimo_dog                   18
Cardigan                     17
Shetland_sheepdog            17
Name: tweet_id, dtype: int64
In [30]:
# Plotting a Bar Chart

# Plot Size 
plt.figure(figsize=(5,5))

# Bar Chart 
most_common.plot(kind= 'bar')

# Plot Annotation
plt.ylabel('Count')
plt.xlabel('Dog Breed')
plt.title('Most Common Breed Type')
plt.show()

Findings:¶

It seems that Golden_retriever is the most common dog in our dataset or tweets, followed by Labrador_retriver, pembroke, chihuahua, pug, chow, etc. next we will be plotting this on a bar char

Dog Breed and Like counts¶

In [229]:
breed_fav = dog_breed[['breed_type', 'favorite_count']] 
breed_fav2 = breed_fav.sort_values('favorite_count', ascending = False).groupby(['breed_type'], sort = False).sum()
breed_fav2.head(20)
Out[229]:
favorite_count
breed_type
Labrador_retriever 898663
Lakeland_terrier 187070
Chihuahua 589311
French_bulldog 408676
Eskimo_dog 227244
English_springer 121013
standard_poodle 95418
golden_retriever 1445911
Chesapeake_Bay_retriever 206969
Italian_greyhound 118775
chow 350249
Pembroke 836976
Pomeranian 247146
Rottweiler 116625
boxer 78329
cocker_spaniel 268058
Samoyed 440004
Saluki 59045
toy_poodle 199536
flat-coated_retriever 116158
In [230]:
breed_fav2.head()
Out[230]:
favorite_count
breed_type
Labrador_retriever 898663
Lakeland_terrier 187070
Chihuahua 589311
French_bulldog 408676
Eskimo_dog 227244
In [31]:
# Plotting a bar Chart showing the relationship between the variable. 

# Plot Size 
plt.figure(figsize=(5,5))

# Bar Chart 
breed_fav3 = dog_breed[['breed_type', 'favorite_count']] 
breed_fav3.head()
breed_fav_group = breed_fav3.groupby('breed_type')['favorite_count'].sum().sort_values(ascending = False).head(20)         
breed_fav_group.plot(kind= 'bar', figsize = (10, 10))

# Plot Annotation 
plt.ylim((0,breed_fav_group.max()))
plt.xlabel('Dog Breed')
plt.ylabel('Counts')
plt.title('Dog Breed VS Favorite Count')
plt.show()

findings :¶

Dogs from the golden retriver breed seems to have the highest number of like or favorite by people. this is followed by the labrador retriver, pembroke, etc.

Dog Breed and retweet counts¶

In [232]:
breed_ret = dog_breed[['breed_type', 'retweet_count']] 
breed_ret2 = breed_ret.sort_values('retweet_count', ascending = False).groupby(['breed_type'], sort = False).sum()
breed_ret2.head(20)
Out[232]:
retweet_count
breed_type
Labrador_retriever 275614
Eskimo_dog 79413
Chihuahua 179936
Lakeland_terrier 58664
English_springer 43836
standard_poodle 37888
French_bulldog 98110
cocker_spaniel 80495
golden_retriever 400209
chow 94171
toy_poodle 70685
Rottweiler 32954
flat-coated_retriever 29815
Chesapeake_Bay_retriever 52295
Pomeranian 81317
boxer 21882
Italian_greyhound 29324
Pembroke 215676
Samoyed 139162
Great_Pyrenees 50264
In [32]:
# Plotting a bar Chart showing the relationship between the variable. 

# Plot Size 
plt.figure(figsize=(5,5))

# Bar Chart 
breed_ret3 = dog_breed[['breed_type', 'retweet_count']] 
breed_ret3.head()
breed_ret_group = breed_ret3.groupby('breed_type')['retweet_count'].sum().sort_values(ascending = False).head(20)         
breed_ret_group.plot(kind= 'bar', figsize = (10, 10))

# Plot Annotation 
plt.ylim((0,breed_ret_group.max()+2000))
plt.xlabel('Dog Breed')
plt.ylabel('Counts')
plt.title('Dog Breed VS Retweet Count')
plt.show()

Findings:¶

There seems to be a correlation between dog that appear the most and the like and retweet count, as the order in which dog breed type appears in the favorite appeared the same way in the retweet section.

Dog Breed and Dog Rating¶

In [234]:
breed_rating = dog_breed[['breed_type', 'rating_numerator']] 
breed_rating2 = breed_rating.sort_values('rating_numerator', ascending = False).groupby(['breed_type'], sort = False).sum()
breed_rating2.head(20)
Out[234]:
rating_numerator
breed_type
black-and-tan_coonhound 21
golden_retriever 1488
Bedlington_terrier 50
bloodhound 73
Pembroke 973
Rottweiler 192
French_bulldog 281
Eskimo_dog 212
Samoyed 440
Pomeranian 414
Irish_setter 47
Gordon_setter 47
standard_poodle 72
Lakeland_terrier 171
Old_English_sheepdog 134
Chihuahua 812
German_shepherd 220
borzoi 85
briard 37
Border_terrier 74
In [34]:
# Plotting a bar Chart showing the relationship between the variable. 

# Plot Size 
plt.figure(figsize=(5,5))

# Bar Chart 
breed_rating3 = dog_breed[['breed_type', 'rating_numerator']] 
breed_rating3.head()
breed_rating_group = breed_rating3.groupby('breed_type')['rating_numerator'].sum().sort_values(ascending = False).head(20)         
breed_rating_group.plot(kind= 'bar', figsize = (10, 10))

# Plot Annotation 
plt.ylim((0,breed_rating_group.max()))
plt.xlabel('Dog Breed')
plt.ylabel('Counts')
plt.title('Dog Breed VS Dog Rating')
plt.show()